I thought Nate Silver was supposed to be smart. It appears he doesn't understand the most basic aspects of probability

He isn't particularly smart, and his stupid predictions are usually wrong.

The relative probability that the Republicans take the House given that they take the Senate is virtually 100% (>99%).

You're starting too many threads is my thought.

rojostreamofconsciousnessrun.com

Sounds like overall site traffic and ad revenues are down. How to fix? Click bait and controversy!

Yeah… it’s pretty obvious these are not independent events.

Rojo… man… you need a vacation.

Nate Silver's predictions are generally correct when a Democrat wins (e.g., Obama) and incorrect when a Republican wins (e.g., Trump). This is because he is a Democrat and produces biased studies more often in favor of Democrats. This is quite pleasing to the left-leaning New York Times, then ESPN, who own his company.

Others have explained it and it's pretty obvious, but your own source has a chart at the bottom titled "Republicans win the Senate, Democrats win the house" with odds "less than 1%" Your own source explains it.

A big part of 538’s accuracy spiel is capturing correlations in their Monte Carlo simulations, e.g., if Michigan goes red, that says something about which way Ohio would, all using the same third-party polling data and historic correlations. If what you say is true (haven’t checked), their model is currently treating the House race as almost perfectly predictable based on the Senate outcome.

That is not exactly ludicrous (though you may disagree): 82% probability with no other information, and near 100% probability conditioned on the Senate going red.

It's possible that Nate Silver doesn't understand basic probability. Alternately, it's possible that you don't. Which is more likely? Let's be nice and call it a coin toss.

The key thing to understand is that the House and Senate races are not independent events. It's not like tossing two coins, where the results of one toss aren't related to the results of the other toss. In this case, the two results are highly correlated. If the Republicans have a good day in the Senate, they're also likely to have a good day in the House, and vice versa. So when 538 runs its model, the runs where the Republicans control the Senate (which is somewhat less likely for them) end up coinciding with runs where they control the House (which is highly likely under any circumstances). In the 18% of runs where Republicans lose the House, they also lose the Senate. Run this past your common sense detector, and you'll see that it passes the sniff test: if something goes wrong enough for the Republicans that they somehow lose the House, it's virtually certain that they'll also lose the Senate.

He is making predictions and Trump was a very atypical candidate, so yeah that didn't fit the models as well. Nate Silver's predictions are based on various models that themselves likely aren't biased, but to the extent they include certain polls etc that may be biased they are biased by that, presumably by using metrics with bias both ways you essentially create a overall probability model that is closer to 'normal.'

The probability that Nate Silver doesn't understand probability is zero.

Correlation my dude.

If GOP takes the Senate then it is virtually 100% they take the House.

If GOP takes the House it is not virtually 100% they take the Senate.

My guess is that they take the House but not the Senate.

Herschel Walker will lose. Dr. Oz will lose. Nevada will flip.

Alan

Yes, you're smarter than both Nate Silver and Malcolm Gladwell. Take the rest of the day off!

What is the deal with Rojo thinking he knows better than everyone else on almost every subject? Narcissism? Mania? Frontal lobotomy?

As the old saying goes Rojo, sometimes it's better to keep your mouth shut and appear stupid than to open it and remove all doubt. :)

In this case, as others have pointed out, the 18% world in which the Democrats win the House is one in which it's virtually certain in 538's model that Democrats also do enough to win the Senate.

I will also point out that the 538 model IS set up to be largely independent from human input, at least until the recalibrate it after the election. There are times when Silver (and others at 538) have questioned the model and they are open on that. Silver has stated that he thinks the model is too bullish on Democrats.

In this case, I personally believe that the model is predicting a bit too high a correlation. There is a world higher than 1% in my view of the Democrats winning the house but Republicans squeaking out victories is PA and either NV or GA but that may be just my personal bias.

He is Ivy league trained.

This is incredibly tough. Silver cracked the code in 2008, and his reputation today largely rests on his correct prediction that Obama would win. Since 2016, though, he's been off, and sometimes horribly so. The clowns on PredictIt are having a field day with his last name.

But...I think it's important to take a step back and remember that his prediction here is based on a model that he has designed, and that we don't know the specifics of that model. A lot of it clearly rests on polling, and...well...the polls have been off for several election cycles. So much so that RealClearPolitics has taken the unprecedented step of *adjusting* polls based on previous error (which I think is highly dubious) to get the likely outcome.

I don't think his model accounts for variables that matter to the average American (inflation, esp. the cost of gas) and he has a clear blindspot for the culture wars. Culture war issues have turned every election post-2016 into something like a collective LARP of "Gangs of New York" where opposing ideologies duke it out to see "who holds sway over the Five Points."

Lastly, he seems to be defining his overall prediction on "possible outcomes" and tallying the number of those outcomes to figure out who is favored. So...in what world is a 56 seat Dem senate even a possible outcome? Or any outcome that has more than 50 Dem seats, for that matter.

As always, it is all about turn out. When more people turn out in any election DEMS win.

Alan